Avalara on July 21 released new research showing that finance leaders are under growing pressure to deploy agentic AI quickly and prove returns, even as governance, accountability, and internal controls lag behind. The report, “Agents of Change,” is based on a survey of more than 1,500 CFOs and senior finance leaders across the US, UK, India, and Australia whose organizations have deployed, piloted, or evaluated AI agents in financial processes during the past year.
Avalara’s report says agentic AI is moving into tax, compliance, financial close, accounts payable, and invoicing, making it part of the operational core rather than a side experiment.
The pressure is nearly universal. Avalara found that 92% of finance leaders feel moderate or significant career pressure to demonstrate ROI from AI agent investments, while half describe that pressure as significant. At the same time, half say their AI agent initiatives have delivered only limited measurable ROI so far, and 71% say deployment pressure is focused primarily on speed.
That creates a difficult operating environment for CFOs. Finance processes are auditable, highly controlled, and hard to unwind when something goes wrong. AI agents that recommend or execute actions in tax, compliance, payments, reporting, or invoice workflows need stronger oversight than generic productivity tools.
Governance Is Falling Behind Deployment
Avalara’s most striking finding is that 76% of surveyed finance leaders say they lack dedicated in-house finance expertise to understand how their AI agents work. Some rely on IT, others rely on vendors, and 16% say no one is currently responsible for it.
The control gaps extend beyond expertise. Avalara found that 30% of respondents have not updated their internal control framework within the past year to account for AI agents taking or recommending actions, while 46% have AI incident response plans that are either untested or still in development.
The report also points to an accountability problem. In the event of a significant AI agent error, 23% of respondents said accountability would be unclear or sit with no one. Another 20% said accountability would sit with the person who deployed or manages the agent, 19% pointed to the team managing it, and 16% pointed to the executive who approved the AI investment.
That ambiguity is especially risky in finance. If an AI agent causes a tax, compliance, payment, reporting, or audit issue, the organization needs to know who approved the workflow, what data the agent used, which controls applied, and how the action can be explained after the fact.
Get Our Free Weekly Newsletter
Finance Leaders Want Controls Built into the System
Avalara’s survey suggests CFOs are not rejecting agentic AI. They are drawing harder lines around where it can act and what proof vendors need to provide.
When asked what would increase confidence in deploying or expanding AI agents across financial environments, respondents pointed to agents operating within the rules, permissions, and controls of existing systems of record, outputs grounded in verified tax, compliance, or financial data, evidence that outputs are tested against known compliance requirements, vendor commitments around accuracy and accountability, and documented audit trails showing what the agent did and why.
The report also found that finance leaders remain cautious about general-purpose LLMs in regulated workflows; 29% said general-purpose LLMs could support regulated workflows only if every output is reviewed by a human, while other respondents said they should be limited to low-risk tasks or are not yet appropriate for decision-making and actions in regulated workflows.
That distinction is important for ERP leaders. Finance AI cannot be treated as a chatbot layer sitting outside the system of record. The value comes when agents operate within approved permissions, use verified finance and compliance data, create audit trails, and escalate exceptions before a financial or regulatory error becomes difficult to reverse.
Sponsor Industry‑Grade Research
What This Means for ERP Insiders
Finance AI needs governance before scale. Agentic AI is moving into workflows where mistakes can affect tax, payments, close, reporting, compliance, and audit evidence. For CFOs, controllers, and ERP finance leaders, the next priority is to define ownership, controls, and escalation paths before agents move from recommendations into actions.
Systems of record will become the control boundary for finance agents. CFOs want agents that operate inside existing permissions, rules, workflows, and audit structures rather than loose automations built around general-purpose models. For ERP teams and finance transformation leaders, the practical test is whether AI agents can work with trusted data and leave a clear record of what happened.
Vendor accountability will become part of AI buying criteria. Avalara’s findings show that finance leaders want proof around auditability, explainability, verified data, compliance testing, and contractual responsibility. For ERP vendors, tax platforms, systems integrators, and finance automation providers, AI credibility will depend less on demos and more on whether customers can defend agent actions to auditors, regulators, and boards.
